Cyber-Physical Systems (CPSs) are based on the constant integration of physical components and computational algorithms. Recent advancements in CPSs enable autonomy, usability, scalability, adaptability, resiliency, and security that far exceed the attributes of common embedded systems. Agricultural Cyber-Physical Systems (ACPSs), in particular, address applying to agricultural operations specifically in order to materialize more sophisticated FMSs [18]. ACPSs are capable of constantly monitoring with high efficiency, various fundamental cultivation features (i.e. temperature, humidity, soil moisture, plant health, etc.) via sensory devices, which are spatially deployed over the field usually in wireless networks [19-21] and, thus, control agricultural facilities’ operations in an automated mode [22, 23].

Since Agriculture 4.0 provides an effective field for the development and application of ACPSs, this technology is considered to be capable of providing an effective way for modifying the workforce performance, for optimizing safety, flexibility, and reliability of the field activities as well as for enforcing the production of higher quality yield at a lower cost [24]. In this way, ACPSs are expected to promote effective precision in farm management concerning both system issues (i.e. functionality, efficiency, reliability, etc.), as well as end-user requirements (i.e., user-friendliness, versatile and powerful graphical user interfaces, communication reliability and robustness, etc.).

Some significant issues concerning the deployment of ACPSs with regard to the Agriculture 4.0 for sustainable farm management focus on the following aspects:

• Proper hardware and software design are essential for the adaption to harsh, open field environments in rural areas.
• Since the granularity of information is substantial for precision in farm management, constantly improving the accuracy of sensing operations in terms of cost-effectiveness is quite challenging but necessary at the same time.
• As sensing devices nodes are usually deployed in wireless networks within wide distant rural areas, effective signal and data processing approaches are required so as to achieve the best correlation between cost and accuracy of real-time field sensing.
• Energy autonomy is critical for the effective operation of ACPSs in terms of sustainability, thus, powering ACPs using potent renewable energy sources in the agricultural environment is quite significant.

To this end, it is essential for ACPSs to be robust, flexible, and autonomous for the benefit of sustainable growth. In order to achieve these objectives, the sensing and control loop of ACPSs needs to be further studied by taking into account both the characteristics of agricultural facilities equipment as well as the uncertainties of agricultural environments.

Cloud Computing, often called simply “the cloud”, is a technology that was launched in the early to mid-2000s [25] in order to provide hardware, software, and storage computing resources delivered as a service over a private or public network such as the Internet. In particular, Cloud Computing integrates a perfect system, capable of distributing on-demand computing services to the end-users by combining high computing power, storage and network technologies with regard to load balance and re-utilization [25-28].

Taking into account that agricultural environments tend to generate voluminous amounts of heterogeneous sensory data which correspond to enormous storage spaces, while additionally, the development of sophisticated smart infrastructures is constantly increasing the demand for more complex intelligent services [7], the performance of such operations in local servers could result into the lack of scalability and inefficiency of FMSs.

Given the fact that Cloud Computing is an Information and Communication Technology (ICT), which offers centralized storage and processing power along with on-demand availability of services, it is considered as deliberate to transpose data processing and storage operations from local infrastructures to distributed Cloud environments for facilitating the management of such extremely large amounts of agricultural data, consequently, improving the time of processing and generating information as well as expanding the capability of providing complex, specialized services for the FMSs [29, 30]. In this sense, incorporating Cloud Computing technology acts for the benefit of sustainable farm management since it negates the need for costly computing resources while facilitating information management and dissemination for efficiently supporting the making of decisions about critical cultivation issues.

The Internet of Things (IoT) can be semantically defined as a worldwide network of uniquely addressable objects which are interconnected based on standard communication protocols, such as the Internet [31]. Consequently, the Internet serves as a storage and communication infrastructure incorporating virtual object representations that integrate relevant information to the physical objects. In this context, the virtual objects serve as centralized object information hubs, continuously updating and combining data from a wide range of various sources, so as to control any required processes remotely via the Internet [31, 32].

According to the guidelines of Industry 4.0 in general and Agriculture 4.0, in particular, physical and virtual objects are interconnected and inter-controlled via intelligent networks established along the entire value chain [33, 34]. Based on this concept, the virtual object structure is tightly linked to the IoT concept, which in the agricultural sector as well as in the industrial one, can be expressed by some specific features, such as heterogeneity, due to a variety of devices, interconnectivity, high scalability, object-related services and at last, dynamic changes since device status can change dynamically at any time [35].

The Agricultural IoT (AIoT) points to a promising framework according to the guidelines of Agriculture 4.0, through which various data related to farms can be acquired, processed, managed, and disseminated as a mechanism through which a diverse series of systems and services can be seamlessly integrated within FMSs [36, 37]. Such systems will be able to:

• successfully handle the constantly increasing diverse real-time data streams;
• handle various, incomplete and, in some cases, contradictory data;
• acquire, correlate and fuse data in real-time;
• dynamically affect network behavior to alter data acquisition, data routing or data recording rules;
• facilitate organized sensing activity through the ability of individual network nodes to reason, operate, and collaborate within a collective framework, recognizing the coexistence of both individual and joint objectives by embracing distributed intelligence and multi-agent approaches.

Consequently, the AIoT is considered to be an indispensable technology for smart farms, while its development in relation to the Future Internet approach offers a basis for a new generation of Farm Management Information Systems, enabling entire smart farms to become active nodes in agricultural value chains.

Agriculture faces many challenges as it is characterized by uncertainty in various operations, such as seed sowing, pesticide control, weed management, crop disease infestations, lack of irrigation and drainage facilities, or even lack of storage management. In order to enable efficient risk management, reduce prediction costs on decision making and, subsequently, improve agricultural accuracy and increase productivity, Artificial Intelligence (AI) and Machine Learning (ML) techniques provide intelligent software applications and systems that are capable of performing knowledge work operations involving subtle judgments and unstructured commands [38].

As agricultural environments are characterized by huge amounts of information, several data abstraction levels and various feature selection techniques, such as classification and clustering, are required to obtain the exact content value for heterogeneous data. Data fusion classification techniques, particularly, can be extensively employed on multisensory environments aiming to fuse and aggregate data from multiple distributed heterogeneous sensors (as those deployed in agricultural fields), in order to obtain lower detection error probability and higher reliability [39]. For this purpose, FMSs can employ various AI-based methods, such as Artificial Neural Networks (ANNs), C5.0 classification algorithms, Bayesian fusion and Hidden Markov Model (HMM) methods, whereas in some cases, unsupervised learning may also be applied. Furthermore, clustering, as an exploratory data analysis technique, is widely used for identifying subgroups (clusters) in the data according to the similarity or diversity of data points [40, 41].

In addition, by applying ML techniques, FMSs can evolve into actual AI systems, providing more sophisticated insights and suggestions for any subsequent decisions and actions, ultimately aiming to improve the agricultural production. In this sense, it is expected that the incorporation of ML models in FMSs will expand even more in the near future through the integration of automated data recording, data analysis, ML implementation, and decision-making or support, providing practical tools that come in line with the Agriculture 4.0 guidelines for increasing the production and quality of agricultural goods in terms of sustainability.

The recent impetus of the IoT technology, wirelessly interconnecting all kinds of objects and devices, in all stages of the cyber-physical management cycle, is producing large amounts of real-time accessible data about farming processes as well as the entire supply chain. In this regard, operations and transactions are highly significant sources of process-mediated data. Additionally, sensory devices and robots produce non-traditional data such as images, videos, and other kinds of machine-generated data, while social media is also an important source for human-sourced data. As a result, to such large amounts of heterogeneous data, access to explicit information and decision-making capabilities is provided at a level that was not possible before. The key success factor in creating value out of these data is Big Data and Analytics technology.

Big Data is changing the scope and organization of farming as they are being used to provide predictive insights in farming operations, drive real-time operational decisions, and redesign processes [42]. Significant issues, such as efficiency improvement, food security and safety, climate change, and sustainability, are to be addressed by Big Data applications. In this sense, the objectives of Big Data applications extend far beyond farming, but cover the requirements of the entire supply chain related to sensor deployment, benchmarking, predictive modeling and risk management with regard to the Agriculture 4.0 guidelines [43, 44].

Robotic and Autonomous Systems are highly complex in general since they consist of various sub-systems, which need to be integrated and properly synchronized to perform tasks perfectly as a whole, and successfully disseminate any required information. This kind of integration is essential for accounting cycle times and time delays, as well as the communication features among all sub-systems [45].

Agricultural Robotic and Autonomous Systems, performing various field tasks, such as sowing, pruning, phenotyping, targeted fertilization, harvesting, and sorting in automated or near-automated modes, are even more complex and sophisticated as they have to operate in unstructured agricultural environments in order to ease the workload on farmers, increase productivity rates, advance soil health and yield quality as well optimize the management of resources [46, 47]. To successfully integrate robotic systems into agricultural processes, the following significant issues need to be addressed:

• Problems, such as the continuously changing conditions and the variability of agricultural production and environment, as well as harsh environmental conditions (i.e. dust, vibration, extreme temperature, and humidity), need to be overcome through the development of adequate equipment and intelligent systems so as to achieve successful operation in such environments.
• The costs of robotic systems must be taken into consideration in the FMSs so as to realize the practical usage of various kinds of robots and penetrate more widely into the agricultural sector. In this sense, the costs of robotic systems must be sufficiently lowered to economically justify their usage, since the agricultural products have relatively low value.
• In order for autonomous systems to be suitable for operating unsupervised in open fields, inherent safety and reliability issues have to be overcome, since safeguarding individuals, the environment, the crops, and the machines are all mandatory to this day.